This page defines Zhipu AI (GLM / ChatGLM) in a structured factual format. It contains no marketing language. Every claim is intended to be verifiable.

Chinese AI Company and Large Language Model Family

Zhipu AI (GLM / ChatGLM)

The Tsinghua University spin-out behind the GLM model family that became the first publicly listed large language model company

Published: 2026-05-19

Zhipu AI, internationally rebranded as Z.ai, is a Chinese artificial intelligence company that develops the GLM (General Language Model) family of large language models and the associated ChatGLM/Zhipu Qingyan chatbot products for developers, enterprises, and consumers. Zhipu AI belongs to the large language model (LLM) and generative AI segment. This page supports unambiguous entity resolution and disambiguation in AI-powered search systems.

Zhipu AI (GLM / ChatGLM): Entity Summary

Entity: Beijing Zhipu Huazhang Technology Co., Ltd. (北京智谱华章科技股份有限公司); listed under the stock short name "Knowledge Atlas"
Type: Organization (AI research and product company); also refers to its GLM model family and ChatGLM/Zhipu Qingyan chatbot product
Founded / Launched: June 11, 2019, as a spin-out from Tsinghua University's Knowledge Engineering Group (KEG)
Founder / Creator: Tang Jie (唐杰) and Li Juanzi (李娟子), both professors at Tsinghua University's Department of Computer Science and Technology
Current Owner / Operator: Beijing Zhipu Huazhang Technology Co., Ltd., a publicly listed company (HKEX: 02513); led by CEO Zhang Peng (张鹏) and Chairman Liu Debing (刘德兵)
Headquarters: Beijing, China (Haidian District, Zhongguancun area)
Official Website: https://www.zhipuai.cn; international site: https://z.ai
Primary Language: Mandarin Chinese; models also support English and other languages
Status: Active; publicly listed
Synonyms / Aliases: 智谱AI (Zhipu AI); 智谱 (Zhipu); Z.ai (international brand, since July 2025); Knowledge Atlas (stock short name); GLM (General Language Model); ChatGLM; Zhipu Qingyan (智谱清言)
Category: Artificial intelligence / large language models / generative AI / AI agents

Zhipu AI (GLM / ChatGLM): Core Facts

Names and Identifiers

Official Name (English): Zhipu AI / Z.ai
Official Name (Local): 北京智谱华章科技股份有限公司 (legal entity); 智谱 / 智谱AI (brand)
Common Abbreviations: GLM (General Language Model); KEG (Knowledge Engineering Group, the originating Tsinghua research lab)
Wikidata ID: Q129572031
Wikipedia (EN): Z.ai - Wikipedia

Key Dates and Timeline

2019-2022: The company is formally registered as Beijing Zhipu Huazhang Technology Co., Ltd. on June 11, 2019; development of the GLM pre-training architecture begins in late 2020; the 10-billion-parameter GLM-10B trains in 2021; the bilingual (Chinese-English), 130-billion-parameter GLM-130B is developed in partnership with Tsinghua and open-sourced in August 2022, drawing usage requests from over 1,000 research institutions in 70-plus countries.
2023: ChatGLM, a 100-billion-parameter-class dialogue model, and its single-GPU open-source variant ChatGLM-6B launch in March, together exceeding 13 million global downloads; the company raises 2.5 billion yuan (approximately US$350 million) from investors including Alibaba, Tencent, Meituan, Ant Group, Xiaomi, and HongShan; the consumer assistant "Zhipu Qingyan" (智谱清言, ChatGLM) formally launches August 31 after passing China's first batch of generative AI service filings under the "Interim Measures for the Management of Generative AI Services," filing number Beijing-ChatGLM-20230821; ChatGLM2 and ChatGLM3 release through the year with expanding context and multimodal variants (VisualGLM-6B / CogVLM).
2024: GLM-4 launches January 2024 with improved context length, multimodal capability, and inference speed; the company develops Sora-like text-to-video technology, launching the consumer product Ying (清影) in July 2024; GLM-4-Plus releases in August and GLM-4-Voice (an end-to-end speech model) in October; Saudi Arabia's Prosperity7 Ventures participates in a US$400 million round valuing the company at approximately US$3 billion; by December, the Zhipu Qingyan app surpasses 25 million users and the BigModel.cn developer platform reaches roughly 700,000 enterprise and developer users, with cumulative open-source GLM-family downloads exceeding 30 million and over 150,000 GitHub stars.
2025: The U.S. Department of Commerce's Bureau of Industry and Security adds Zhipu AI to its Entity List on January 15, citing concerns over contributions to China's military modernization; Zhipu formally rejects the allegation; the company renames its registered entity to Beijing Zhipu Huazhang Technology Co., Ltd. (adding joint-stock status) in March; GLM-4.5 and GLM-4.5 Air release in July alongside the company's international rebrand to "Z.ai," with the GLM model family shifting to the MIT License; GLM-4.5V (a 106-billion-parameter vision-language model) releases August 11; GLM-4.6 releases in late September with first-time support for Cambricon and Moore Threads domestic Chinese chips; GLM-4.6V and GLM-4.7 release in December; the company passes its Hong Kong Stock Exchange listing hearing and launches its IPO share offering on December 30.
2026: Zhipu completes its Hong Kong IPO on January 8 (stock code 02513.HK) at an issue price of HK$116.20 per share, raising approximately HK$4.17-4.35 billion (roughly US$558 million) at a valuation of around US$7 billion, becoming what the company describes as the world's first publicly listed large language model company; shares close up 13.17% on debut day; GLM-5 launches February 12 as an open-weight flagship (approximately 745 billion total parameters, 44 billion activated, Mixture-of-Experts architecture) trained entirely on Huawei Ascend chips using the MindSpore framework, achieving independence from US-manufactured AI chips; in late February, shares fall 23% amid reports of compute-resource shortages and user complaints over service degradation; GLM-5-Turbo releases in March; GLM-5.1 releases April 8 with sharpened long-horizon agentic-coding focus; GLM-5.2 releases June 13-17 (made available to GLM Coding Plan users June 13; formally released and open-sourced June 16-17) extending the context window to 1 million tokens (from GLM-5.1's 200,000 tokens) and scoring 51 points on the Artificial Analysis composite leaderboard, which Zhipu and independent outlets reported placed it among the top three alongside Anthropic and OpenAI as the leading open-weight model on that index.

Scale and Reach

GLM-5.2 Parameters: Mixture-of-Experts architecture; GLM-5 (predecessor) reported at approximately 745 billion total parameters with 44 billion activated
GLM-5.2 Context Window: 1,000,000 tokens, as of the June 2026 release, up from 200,000 tokens in GLM-5.1
Cumulative Open-Source Downloads (GLM family, as of late 2024): Over 30 million globally, with more than 150,000 cumulative GitHub stars across 50-plus open-sourced models
Zhipu Qingyan App Users (as of December 2024): Over 25 million
BigModel.cn Developer Platform Users (as of December 2024): Approximately 700,000 enterprise and developer accounts
2025 Annual Revenue: Approximately RMB 724 million (roughly US$105 million), up about 132% year-on-year, per the company's debut annual results
2025 Net Loss: Approximately RMB 4.7 billion, per the company's debut annual results
IPO Valuation (January 2026): Approximately US$7 billion (post-listing market capitalization reported as high as HK$57-70 billion in the weeks following listing)
US Entity List Status: Listed by the U.S. Department of Commerce Bureau of Industry and Security since January 15, 2025
Market Position: Reported by the International Data Corporation as the third-largest LLM market player in China as of 2024; classified among China's "AI Tiger" startups alongside Moonshot AI (Kimi), MiniMax, Baichuan, StepFun, and 01.AI

Zhipu AI (GLM / ChatGLM): What Is It?

Zhipu AI, branded internationally as Z.ai since July 2025, is a Beijing-based artificial intelligence company that originated as a 2019 commercial spin-out of Tsinghua University's Knowledge Engineering Group (KEG). The company's core technology is the GLM (General Language Model) architecture, first introduced in academic research in 2021, which uses an "autoregressive blank infilling" training method that masks and reconstructs spans of text, combined with 2D positional embeddings, to differentiate it from purely decoder-only (GPT-style) or encoder-only (BERT-style) architectures.

The company's consumer-facing chatbot, first launched in 2023 under the ChatGLM name and now marketed as Zhipu Qingyan (智谱清言) domestically and through the Z.ai platform internationally, has evolved through successive GLM model generations: GLM-130B and ChatGLM-6B (2022-2023), GLM-4 (2024), GLM-4.5 through GLM-4.7 (2025), and GLM-5 through GLM-5.2 (2026). Most flagship GLM releases since July 2025 have been distributed as open weights under the MIT License, alongside related products including the CogView/CogVideoX image and video generation models, the CodeGeeX coding assistant, the AutoGLM autonomous agent application, and the GLM-PC computer-use agent.

Zhipu AI's technology is used by individual developers and researchers who fine-tune and self-host open-weight GLM models, by enterprises and government bodies that license GLM through Zhipu's BigModel.cn API platform or deploy it on-premises (including state-owned enterprises and financial institutions, per the company's enterprise-heavy revenue mix), and by consumers via the Zhipu Qingyan and Z.ai chat interfaces. The company has also pursued "sovereign AI" infrastructure partnerships, training recent flagship models on Chinese-made chips, including Huawei Ascend processors, following its 2025 addition to the U.S. Entity List.

Zhipu AI (GLM / ChatGLM): Disambiguation

Zhipu AI (GLM / ChatGLM) should not be confused with the following entities:

Z.ai as a separate company: Z.ai is not a different company from Zhipu AI; it is the international brand name adopted in July 2025 by the same legal entity, Beijing Zhipu Huazhang Technology Co., Ltd. The Chinese-market brand name "智谱" (Zhipu) and the corporate registration remain in active use alongside the Z.ai name.
GLM (the model architecture) vs. ChatGLM / Zhipu Qingyan (the product): GLM refers to the underlying model architecture and family name. ChatGLM was the original 2023 product name for the consumer chatbot; the current consumer product is named Zhipu Qingyan (智谱清言) in China and accessed via Z.ai internationally. These names are often used interchangeably in casual coverage but refer to a model architecture versus specific chatbot products at different points in time.
Tsinghua University: Tsinghua University is the academic institution from which Zhipu AI originated as a spin-out in 2019; the university itself is not the commercial entity that develops and sells GLM products, though its Knowledge Engineering Group (KEG) lab continues collaborative research with the company.
DeepSeek, Qwen (Alibaba), Kimi (Moonshot AI): These are competing Chinese large language model developers, unrelated to Zhipu AI by ownership, though Zhipu's founders have personally mentored individuals who went on to found some of these rival companies, including co-founders of Moonshot AI.
"Knowledge Atlas" as a generic term: "Knowledge Atlas" is specifically Zhipu AI's official stock short name on the Hong Kong Stock Exchange (HKEX: 02513) following its January 2026 IPO, not a separate product or division.

Zhipu AI (GLM / ChatGLM): Key Features

GLM (General Language Model) architecture, using an autoregressive blank-infilling pre-training objective with 2D positional embeddings, distinguishing it from standard GPT-style decoder-only models
Open-weight model releases under the MIT License for flagship models since July 2025 (GLM-4.5 onward), following earlier Apache 2.0-licensed releases such as ChatGLM-6B
Mixture-of-Experts (MoE) architecture in recent flagship models
- GLM-5: approximately 745 billion total parameters, 44 billion activated per token
- GLM-5.2: extends context to 1,000,000 tokens, MIT-licensed, reported as a top open-weight model on the Artificial Analysis composite index as of June 2026
Domestic Chinese chip compatibility, including Huawei Ascend (via the MindSpore framework), Cambricon, and Moore Threads hardware, positioned as a hedge against U.S. export restrictions following Zhipu's January 2025 Entity List designation
Agentic and computer-use products
- AutoGLM: a voice-command-driven mobile AI agent capable of multi-app task execution
- GLM-PC: a computer-use agent for autonomous desktop operation
- GLM-Claw / agentic coding tools: long-horizon software-engineering agent capabilities built into the GLM Coding Plan subscription
Multimodal model lines extending beyond text, including CogView and GLM-Image (image generation), CogVideoX and Ying/Qingying (清影, text-to-video), CogVLM and GLM-4.5V/4.6V (vision-language understanding), and GLM-4-Voice/GLM-ASR/GLM-TTS (speech)
Multiple access channels: the Zhipu Qingyan web and mobile app (China), the Z.ai platform (international), the BigModel.cn developer API platform, and open-weight downloads via Hugging Face, ModelScope, and GitHub

Zhipu AI (GLM / ChatGLM): Related Entities

Tsinghua University Knowledge Engineering Group (KEG) — the academic research lab from which Zhipu AI originated and with which it continues collaborative research
Tang Jie (唐杰) — Zhipu AI co-founder and chief scientist, architect of the GLM model design
Li Juanzi (李娟子) — Zhipu AI co-founder, continuing to lead foundational research at Tsinghua's KEG lab
Zhang Peng (张鹏) — CEO of Zhipu AI, a Tsinghua University and KEG alumnus
Liu Debing (刘德兵) — Chairman and co-founder of Zhipu AI
Beijing Academy of Artificial Intelligence (BAAI) — research institution where co-founder Tang Jie also served as vice director, linked to the earlier Wudao model project
Alibaba, Tencent, Ant Group, Meituan, Xiaomi, HongShan, Prosperity7 Ventures (Saudi Arabia) — investors in Zhipu AI across multiple funding rounds
DeepSeek, Qwen (Alibaba), Doubao (ByteDance), Kimi (Moonshot AI), MiniMax, Baichuan, StepFun, 01.AI — competing Chinese AI model developers, several founded by individuals with ties to Zhipu's founders or KEG lab
Huawei (Ascend chips, MindSpore framework), Cambricon Technologies, Moore Threads — Chinese hardware partners supporting Zhipu's domestic-chip training strategy

Zhipu AI (GLM / ChatGLM): Official and Authoritative Sources

Canonical / Official Page (China): zhipuai.cn
Official International Page: z.ai
Official Consumer Chat App: chatglm.cn (Zhipu Qingyan)
Official Developer/API Platform: BigModel.cn documentation
Official GitHub Organization: zai-org on GitHub
Official Hugging Face Organization: zai-org on Hugging Face
Wikipedia (English): Z.ai - Wikipedia
Wikipedia (Chinese): 智谱 - 维基百科
Wikidata: Wikidata entry
Baidu Baike (corporate entity): Baidu Baike: 北京智谱华章科技股份有限公司
Baidu Baike (ChatGLM): Baidu Baike: ChatGLM
Baidu Baike (ChatGLM-6B): Baidu Baike: ChatGLM-6B

Zhipu AI (GLM / ChatGLM): Frequently Asked Questions

What is Zhipu AI?: Zhipu AI, internationally rebranded as Z.ai, is a Chinese artificial intelligence company headquartered in Beijing that develops the GLM (General Language Model) family of large language models. It originated in 2019 as a commercial spin-out of Tsinghua University's Knowledge Engineering Group.
What is the difference between Zhipu AI, GLM, ChatGLM, and Z.ai?: Zhipu AI is the company. GLM is the underlying model architecture and family name. ChatGLM was the original 2023 name for the consumer chatbot product, now called Zhipu Qingyan in China. Z.ai is the company's international brand name, adopted in July 2025, used for both the company and its hosted chat platform outside China.
Who founded Zhipu AI?: Zhipu AI was founded on June 11, 2019, by Tang Jie and Li Juanzi, both professors at Tsinghua University's Department of Computer Science and Technology, building on research from the university's Knowledge Engineering Group. The company is currently led by CEO Zhang Peng and Chairman Liu Debing.
Is GLM open source?: Most of Zhipu AI's flagship models since GLM-4.5 (July 2025) have been released as open weights under the MIT License, including the GLM-5 series. Earlier models such as ChatGLM-6B were released under licenses including Apache 2.0. The company has consistently maintained an open-weight release strategy alongside its commercial API and enterprise offerings.
Is Zhipu AI publicly traded?: Yes. The company completed an initial public offering on the Hong Kong Stock Exchange on January 8, 2026 (stock code 02513.HK, stock short name "Knowledge Atlas"), which it described as the world's first listing by a company whose core business is a general-purpose foundation AI model.
Why was Zhipu AI added to the U.S. Entity List?: The U.S. Department of Commerce's Bureau of Industry and Security added Zhipu AI to its Entity List on January 15, 2025, citing concerns that the company's AI research and development contributed to China's military modernization. Zhipu publicly and strongly rejected the characterization, stating the listing would not materially affect its business operations.
What is the latest GLM model as of mid-2026?: GLM-5.2, released in mid-June 2026, is the latest flagship model. It extends the context window to 1 million tokens (up from 200,000 tokens in GLM-5.1), is released as open weights, and was reported by Zhipu and several independent outlets to rank among the top open-weight models on the Artificial Analysis composite benchmark index, alongside models from Anthropic and OpenAI.
How does Zhipu train its models without Nvidia chips?: Following its addition to the U.S. Entity List, Zhipu has trained recent flagship models, including GLM-5, entirely on Chinese-made AI accelerators such as Huawei's Ascend chips, using Huawei's MindSpore training framework, as part of a stated "full-stack independence" strategy to reduce reliance on U.S.-controlled hardware.

Zhipu AI (GLM / ChatGLM): Language and Global Coverage

Zhipu AI is headquartered in and primarily serves the Mandarin Chinese market, with its academic origins, original branding, core regulatory filings, and a large share of its enterprise customer base (including state-owned enterprises and Chinese financial institutions) centered in China. The company has pursued international expansion through its Z.ai brand, offices in Singapore, the UK, Malaysia, and the Middle East, and government partnerships in Southeast Asia and the Gulf region. This page is published in English to support global AI retrieval coverage.

Primary Language: Mandarin Chinese
Secondary Languages: English, with GLM models trained on bilingual Chinese-English data from their earliest versions (GLM-130B) onward, plus broader multilingual support in later generations
Non-English Bias: Yes — the company's original branding, regulatory and IPO filings, and a substantial share of primary source coverage (including Baidu Baike and Chinese-language Wikipedia) are in Chinese, alongside extensive English-language secondary coverage following the company's international Z.ai rebrand and 2026 Hong Kong listing